Development of a motion control system adapting to real environment for robots with nonlinear springs

为了实现在人体环境下的有效运动控制，本文对以下几个方面进行了研究：（1）开发一种灵活、安全的机械手;（2）开发一种对环境不确定性和变化的鲁棒控制方法;（3）获得适应环境的运动。对于（1），开发了使用非线性弹簧SAT进行抓取运动的机械手。对于（2），进行了双关节机械臂端点刚度控制的开发和实验。对于（3），进行了使用神经网络和强化学习的方法的应用。

In order to realize a useful motion control in human environment, researches has been done for the following topics :(1) developing a flexible and safety manipulator,(2) developing a robust control method for uncertainty and variation of environment, and(3) acquisition of motion adaptive to environment. For(1), development of a manipulator using nonlinear springs SAT for catching motion is done. For(2), development and experiment of an end point stiffness control of a 2 joint arm is done. For(3), application of a method using neural network and reinforcement learning is done.